Your AI agents can now control your Google Home devices

Sarah Perez at TechCrunch reports Google rolled out early access to a Model Context Protocol server for Google Home. Any MCP-capable agent—Claude, Hermes, OpenClaw, ChatGPT, Google Antigravity—can work with smart-home devices and event history.

Natural-language control of cameras, activity, lights, and custom dashboards stops being a Google Assistant monopoly and becomes an agent marketplace. Setup still requires a Google Cloud project, Home MCP config, sign-in, and granted permissions.

That is useful when it works. It is also a new attack surface sitting next to your doorbell and thermostat.

We’re for agents that do real chores—if the permission model is boringly strict.

Google’s Home MCP early access, as TechCrunch describes it, lets people review camera summaries, monitor activity, control connected devices, and build dashboards through whichever MCP agent they already use. Nest doorbells and thermostats plus Works-with-Google-Home and Matter devices are in scope. Google already uses MCP elsewhere, including Maps; this extends the pattern into the physical home.

The useful read: interoperability beats yet another proprietary voice island. If Claude or ChatGPT can flip lights without a second app religion, that’s progress people feel.

The harm path is obvious. An agent with camera history and device control is a high-value target. Early access via Cloud project setup will filter to technical users first—good for caution, bad if the eventual consumer path softens consent. I’m watching how revocation works when an agent misfires, and whether Google ships clear audit logs of every agent action.

This is the smart home becoming an API. We’re glad when APIs are open. We insist when homes are involved that open doesn’t mean careless.

Model Context Protocol is the emerging standard for giving AI agents structured tool access. Google Home already spans Nest hardware and Matter third-party devices; MCP is the bridge to third-party chat agents.